JP2020095448A - Data processing system, method and program - Google Patents

Abstract

To provide a data processing system, a method, and a program that provide a technique for effectively analyzing an action of a person.SOLUTION: A data processing system, which performs processing of data on an action of a person, includes: a data management unit that stores action data in which a measured value obtained by measuring an action of a predetermined action item by a person is accumulated; a coloration rule that defines a color corresponding to the measured value for each action item; an image output unit that generates, using an arrangement rule that defines where to place a region colored by the coloration rule for each action item, for each person, an action image in which a region colored in a color corresponding to the measured value of action item included in action data for a predetermined period is arranged; and an image analysis unit that learns an action image of a person and correct answer information indicating a property and/or a state of the person, and estimates, based on the learning result, a property and/or a state of the person from the action image of the person.SELECTED DRAWING: Figure 1

Description

The present invention relates to a technique for processing data regarding human behavior.

The decrease in the working population accompanying the declining birthrate and aging population has become a real problem. Organizations such as companies are required to continuously secure human resources and effectively utilize the human resources in order to maintain and improve productivity and competitiveness.

In order to realize it, it is expected to utilize IT (Information Technology) technology. And effective techniques have been realized in some fields.

For example, IT technology has been promoted to reduce or substitute a part of the work of human resources (see Patent Document 1). For example, it can be expected that the work of human resources is reduced by RPA (Robotic Process Automation) or that the work of human resources is performed by the RPA instead.

JP, 2018-022257, A

Furthermore, it is desired to analyze the behavior of a person by IT technology in various fields and use the analysis result. For example, in an organization such as a company, it is desired that the behavior of the employee be analyzed and the analysis result be used so that the ability of the employee can be effectively utilized. Specifically, it is expected that employees will be able to control leave and leave, work with high productivity, and allocate human resources to improve productivity.

However, since human behavior data contains a lot of noise data and the data is highly diverse, it is often the case that an effective analysis result cannot be obtained. For example, noise may be caught as a feature amount in multiple regression analysis. Also, in machine learning, noise may be overlearned. It is difficult to improve the accuracy of the prediction algorithm because a number of factors are complicatedly related to employee leave, employee turnover, high productivity, and so on.

An object of the present invention is to provide a technique for effectively analyzing human behavior.

A data processing system according to one embodiment of the present invention is a data processing system for processing data relating to a person's action, and is data for storing action data accumulating measurement values obtained by measuring actions of a predetermined action item by a person. Using a management unit, a coloring rule that defines a color corresponding to the measurement value for each action item, and an arrangement rule that determines where in the image the region colored by the coloring rule for each action item is arranged, An image output unit that generates an action image in which an area colored in a color corresponding to the measurement value of each action item included in the action data for a predetermined period is arranged for each person, the action image of the person, the property of the person, and Alternatively, the image analysis unit that learns correct answer information indicating a state and estimates the property and/or state of the person from the action image of the person based on the learning result.

It is possible to support the effective use of the ability of human resources.

FIG. 1 is a diagram showing a configuration example of a data processing system according to a first embodiment. It is a figure which shows the structural example of a work management apparatus. It is a figure which shows the structural example of an image output device. It is a figure which shows the structural example of an image analysis apparatus. It is a figure which shows the structural example of a work day table. It is a figure which shows the structural example of a coloring rule. It is a figure which shows the structural example of arrangement|positioning rule. It is a figure which shows an example of action image data. It is a schematic diagram which shows an example of an action image. It is a flow chart which shows an example of image re-learning processing. It is a figure which shows the example of a display of display data. It is a figure which shows an example of the hardware constitutions of the apparatus which concerns on this indication. It is a figure for demonstrating the information collection by a work management apparatus. It is a figure which shows an example of action data. FIG. 9 is a diagram showing an example of a priority table according to the second embodiment. 9 is a schematic diagram showing an example of an action image according to Example 2. FIG.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a diagram illustrating a configuration example of the data processing system according to the first embodiment.

The data processing system 100 is a system for processing data related to human behavior. As shown in FIG. 1, the data processing system 100 includes a work management apparatus 101 that is an example of a data management unit, an image output apparatus 201 that is an example of an image output section, and an image analysis apparatus 301 that is an example of an image analysis section. And a terminal device 801. These devices 101, 201, 301, 801 are connected to a network 901 such as a wired LAN (Local Area Network) and/or a wireless LAN, and can send and receive data to and from each other. Each of these devices 101, 201, 301 and 801 may be the computer device shown in FIG. Further, at least two of these devices 101, 201, 301 may be configured as one computer device.

FIG. 2 is a diagram showing a configuration example of the work management apparatus 101.

As illustrated in FIG. 2, the work management apparatus 101 includes a work day acquisition processing unit 102. The work day acquisition processing unit 102 acquires, for example, a work day table 103, which is an example of behavior data, from a predetermined DB (DataBase) and stores the work day table 103 in a storage device (see FIG. 12 ).

The action data is data in which measurement values obtained by measuring actions of a predetermined action item by a person are accumulated. In the work daily table 103, which is an example of action data, an example of an action item is “attendance time”, and an example of a measured value corresponding to the action item “attendance time” is a time at which an employee actually goes to work. The details of the work day table 103 will be described later (see FIG. 5).

FIG. 3 is a diagram showing a configuration example of the image output device 201.

As shown in FIG. 3, the image output device 201 includes an action image generation processing unit 202. The action image generation processing unit 202 generates an action image 203 for each person (employee) from the work daily table 103 using an image generation rule 206 including a coloring rule 204 and an arrangement rule 205.

The coloring rule 204 defines a color corresponding to a measured value for each action item. The arrangement rule 205 defines for each action item where in the image the area (mass) colored by the coloring rule 204 is to be arranged. The details of the coloring rule 204 and the arrangement rule 205 will be described later (see FIG. 6).

The action image 203 is an arrangement of areas colored in colors corresponding to the measured values of each action item included in the action data of a predetermined period. The details of the action image 203 will be described later (see FIGS. 8 and 9).

FIG. 4 is a diagram showing a configuration example of the image analysis device 301.

As shown in FIG. 4, the image analysis device 301 includes an image input processing unit 302, an image learning processing unit 303, an image classification processing unit 304, and a result output processing unit 305.

The image input processing unit 302 receives an input of the action image 203 generated by the image output device 201.

The image learning processing unit 303 learns a large number of action images 203 and generates an estimation algorithm for classifying the action images 203 into groups having image-wise common features. For example, the image learning processing unit 303 learns a large amount of action images 203 by deep learning and generates a neural network model corresponding to an estimation algorithm. As described above, the deep learning in which the noise is suppressed by the imaging makes it possible to accurately analyze the complex human behavior.

Further, the image learning processing unit 303 associates each group with correct answer information. For example, the administrator associates the correct answer information indicating the nature and/or state of the people (that is, the person's action pattern) corresponding to the action image 203 belonging to the same group with the group. In this way, the behavior of a person can be effectively analyzed by imaging the data of the behavior of the person and processing the image to estimate the nature and/or state of the person.

The image classification processing unit 304 uses the estimation algorithm generated by the image learning processing unit 303 to classify the behavior image 203 of the subject into any of the groups. Then, the image classification processing unit 304 generates the correct answer information associated with the group as the classification result 306. That is, the image classification processing unit 304 estimates the behavior pattern of the target person using the estimation algorithm. In this way, the behavior of the target person can be effectively analyzed by estimating the property and/or state of the target person by using the estimation algorithm learned by imaging the data of the human behavior.

The result output processing unit 305 is an example of an information presenting unit, and generates display data 307 from the classification result 306 generated by the image classification processing unit 304 and displays it on the display device (see FIG. 12). As a result, it is possible to present the nature and/or condition of the target person and the utilization policy for it, so that it is possible to support the effective utilization of human ability.

Further, the result output processing unit 305 may display the behavior image 203 of the subject on the display device. This allows the administrator to intuitively understand the nature and/or state of the target person from the action image 203.

FIG. 5 is a diagram showing a configuration example of the work day table 103.

As shown in FIG. 5, the work day table 103 has, as data items, an employee ID, a date, an attendance time, and a departure time. For example, the first line in FIG. 5 indicates that “employee A” has attended “2018/07/01”, “08:25”, and retired at “18:05”.

FIG. 6 is a diagram showing a configuration example of the coloring rule 204.

The coloring rule 204 defines, for example, as for action items related to working hours, areas of the same action items have the same hue, and the darker the color tone becomes as the working time becomes longer. By defining the coloring rule 204 in this way, a dark area appears in the work image of the worker who spends a lot of time working, so that the length of the working time of the worker can be intuitively grasped. ..

The table in the upper part of FIG. 6 is an example in which a correspondence relationship among action items, color arrangement types, and hues is determined.

The first row of the table in the upper part of FIG. 6 defines that the color arrangement type “shade 1” and the hue “blue” are used for the action item “start time”. Here, as shown in the first row of the table in the lower part of FIG. 6, “shading 1” is the shading “dark” in the range “minimum value to less than 10%” and the range “10% or more to less than 60%. It is defined that the light and shade "slightly dark" is used for "", the light and shade "slightly light" is used for the range "60% or more and less than 90%", and the light and light "light" is used for the range "90% or more to maximum value". That is, according to the coloring rule 204, in the action image 203, the square of the action item “starting time” is colored with a darker color as the time is earlier.

The second row of the table in the upper part of FIG. 6 defines that the color scheme type “shade 2” and the hue “green” are used for the action item “finishing time”. Here, "shading 2" means shading "light" in the range "minimum value to less than 10%" and "range 10% to less than 60%" as shown in the second row of the lower table of FIG. It is defined that the light and shade is "slightly light", the light and shade is "slightly dark" in the range "60% or more and less than 90%", and the light and dark "dark" is used in the range "90% or more-maximum value". That is, according to this coloring rule 204, in the action image 203, the cell of the action item “working time” is colored with a darker color as the time is later.

The fourth row of the table in the upper part of FIG. 6 uses the color arrangement type “color classification 1” and the hues “1: blue, 2: green, 3: yellow, 4: red” for the action item “health management time”. Determine that. Here, as shown in the third row of the table in the lower part of FIG. 6, “color classification 1” means the hue “color 1” in the range “0 or less” and the hue “color 1” in the range “0 or more and less than 15”. It is defined that the hue “color 3” is used for the range “15 or more and less than 45” and the hue “color 4” is used for the range “45 or more”.

FIG. 7 is a diagram showing a configuration example of the arrangement rule 205.

As shown in FIG. 7, the arrangement rule 205 is information in which a daily action item is assigned to each of a plurality of cells (areas) divided vertically and horizontally. For example, as shown in FIG. 7, the placement rule 205 assigns daily action items to a plurality of cells such that dates are arranged in either the vertical direction or the horizontal direction and action items are arranged in the other direction. Determine. By setting the arrangement rule 205 in this way, the measured values of the same item for each date are lined up vertically or horizontally in a line, so that the similarity of actions is likely to appear in the feature amount of the image.

Further, as shown in FIGS. 6 and 7, the action image 203 can be easily generated by coloring the square with the color corresponding to the measurement value of the action item according to the coloring rule 204 and the arrangement rule 205.

FIG. 8 is a diagram showing an example of the action image data 213.

The action image generation processing unit 202 generates the action image data 213 by arranging dates in the vertical direction and action items in the horizontal direction according to the arrangement rule 205, as shown in FIG. Then, the action image generation processing unit 202 associates the value of the action item in the action image data 213 with the hue and shade according to the coloring rule 204, as shown in FIG. For example, the action image generation processing unit 202 associates the hue “blue” and the shade “light” with the start time “08:56” of “2018/04/02” according to the coloring rule 204 shown in FIG.

FIG. 9 is a diagram showing an example of the action image 203.

The action image generation processing unit 202 generates an action image 203 as shown in FIG. 9 based on the action image data shown in FIG. For example, the cell P of the action image 203 shown in FIG. 9 corresponds to the start time “08:56” of “2018/04/02” of the action image data shown in FIG. Therefore, the action image generation processing unit 202 colors the square P of the action image 203 with light blue. The action image generation processing unit 202 similarly colors the corresponding squares in the action image 203 for the closing time and the health care time. Thereby, the start time, the end time, and the health management time can be reflected in the action image 203.

The image learning processing unit 303 generates an estimation algorithm for classifying into a group having image-wise common features, based on the thus colored action image 203.

For example, when the coloring rule 204 in FIG. 6 is followed, the action image 203 of the employee whose working time and leaving time every day are on time becomes pale in color. Employees who perform such work are considered to be suitable for management work, for example. Therefore, the correct answer information, which is suitable for management work, is associated with the group A, which has a light color as a common feature in terms of image. In this case, the image classification processing unit 304 may determine that the target person corresponding to the action image 203 classified into the group A may be suitable for the management work.

For example, when the coloring rule 204 of FIG. 6 is followed, the action image 203 of the employee whose working time every day is earlier than the regular time and/or whose daily leaving time is later than the regular time is dark in color as a whole. It is considered highly necessary for employees who perform such work to follow up on their health. Therefore, the correct answer information indicating that it is necessary to follow the health aspect is associated with the group B, which has a dark color as a common feature in terms of image. In this case, the image classification processing unit 304 may determine that the target person corresponding to the action image 203 classified into the B group needs to follow the health aspect.

It should be noted that the image classification processing unit 304 determines that the normal behavioral image 203 of the subject is classified into the group A, but the recent behavioral image 203 is classified into the group B. It may be determined that it is necessary to follow the face.

In addition, the image classification processing unit 304 associates the correct answer information that the stress resistance is high with the B group, and determines that the target person corresponding to the action image 203 classified into the B group by the image classification processing unit 304 has the stress resistance. It may be determined that there is a high possibility.

For example, when the coloring rule 204 of FIG. 6 is followed, the action image 203 of the employee who is working flexibly has a mottled (separate) color as a whole. Employees who perform such work are considered to be more suitable for spot correspondence of a specific project rather than routine work, for example. Therefore, the correct answer information that it is suitable for spot correspondence is associated with the group C which has a mottled pattern as a common feature in terms of image. In this case, the image classification processing unit 304 may determine that the target person corresponding to the action image 203 classified into the C group may be suitable for spot correspondence.

FIG. 10 is a flowchart showing an example of the image re-learning process.

The image learning processing unit 303 generates an estimation algorithm from the large number of action images 203 as described above. However, when the classification result 306 by the estimation algorithm is significantly different from the assumption of the administrator, or when the behavior data having a new tendency is increased, the estimation algorithm needs to be adjusted. In such a case, the image re-learning process shown in FIG. 10 may be executed. Next, the image learning process shown in FIG. 10 will be described.

First, the work day acquisition processing unit 102 acquires the work day table 103 in the work day acquisition process (S101).

Next, the action image generation processing unit 202 generates the action image 203 from the work day table 103 in the action image generation process (S102).

Next, the image learning processing unit 303 generates an estimation algorithm based on the large number of action images 203 in the image learning process (S103).

Next, in the image classification process, the image classification processing unit 304 classifies the action image 203 into any group using the estimation algorithm (S104).

The administrator determines whether the tendency of the classification result 306 of S104 is correct (S105).

When the administrator determines that the tendency of the classification result 306 is correct (S105: YES), the flow ends this processing.

When the administrator determines that the tendency of the classification result 306 is not correct (S105: NO), the administrator then decides what to adjust (S106).

When the administrator determines to adjust the image generation rule 206 (S106: image generation rule), the flow returns to S102, and the action image generation process is performed using the action generation rule different from the previous one. As a result, the action image 203 different from the previous one is generated.

If the administrator decides to adjust the learning data (S106: learning data), the flow returns to S103 to generate a group and estimation algorithm different from the previous one. Thereby, the action image 203 can be classified into a group different from the last time.

FIG. 11 is a diagram showing a display example of the display data 307.

The result output processing unit 305 generates the display data 307 based on the classification result 306, and displays the display data 307 on the display device as illustrated in FIG. 11.

For example, regarding health care, using an estimation algorithm that classifies into three groups of "need care", "requiring attention", and "no problem", behavior images of each employee in a department in March 2018 As a result of classifying 203, it is assumed that the behavior image 203 of one employee is classified into the "need care" group and the behavior image 203 of two people is classified into the "need attention" group. In this case, as illustrated in FIG. 11, the result output processing unit 305 determines that one employee is “needs care” and two employees are “needs attention” as the March prediction result. indicate.

As a result, the manager of the department or the personnel department can take care of the employee before the health of the employee deteriorates.

FIG. 12 is a diagram illustrating an example of the hardware configuration of the device according to the present disclosure.

As illustrated in FIG. 12, the work management apparatus 101, the image output apparatus 201, the image analysis apparatus 301, and the terminal device 801 according to the present disclosure include an internal bus 1001, a processor 1002 connected to the internal bus 1001, and a main unit. A memory 1003, a storage device 1004, a communication device 1005, an input device 1006, and a display device 1007 are provided.

The processor 1002 is, for example, a CPU (Central Processing Unit) or the like. The main memory 1003 is, for example, a DRAM (Dynamic Random Access Memory) or the like. The storage device 1004 is, for example, an SSD (Solid State Drive) or an HDD (Hard Disk Drive). The communication device 1005 is, for example, an Ethernet (registered trademark) board, a wireless LAN board, or the like. The input device 1006 is, for example, a keyboard and a mouse. The display device 1007 is, for example, a liquid crystal display or the like.

The functions of the work management apparatus 101, the image output apparatus 201, and the image analysis apparatus 301 described above may be realized by the processor 1002 executing a computer program in the storage device 1004 and/or the main memory 1003.

In this embodiment, the action items are simplified by focusing on the start time, end time, and health care time. However, the present invention is not limited to this, and various other items regarding actions may be used as action items.

FIG. 13 is a diagram for explaining information collection by the work management apparatus 101. FIG. 14 is a diagram showing an example of action data.

As illustrated in FIG. 13, the work management apparatus 101 acquires various types of information regarding behavior data from various types of DBs via various types of information integration apparatuses. Then, the work management apparatus 101 generates behavior data in which measured values corresponding to various behavior items on each date are entered for each employee from the acquired various information as shown in FIG. Store. As shown in FIG. 14, the action data may include measurement values relating to various work such as vacation classification, midnight time, and business trip time in addition to the above-mentioned start time, end time, and health management time. ..

In the first embodiment, in the action image 203, an example in which the same size area is assigned to all action items is shown, but the present invention is not limited to this. In the second embodiment, an example is shown in which priorities are given to action items and the area of each action item in the action image 203 has a size corresponding to the priority.

FIG. 15 is a diagram showing an example of the priority table. FIG. 16 is a schematic diagram illustrating an example of the action image 403 according to the second embodiment.

The priority table 401 is a table for setting priorities for action items. In the priority table 401 illustrated in FIG. 15, “1” having the highest priority is set for the start time and the end time of the work, and “3” having the lowest priority is set for the health care time. In this case, the action image generation processing unit 202, when generating the action image 403, as shown in FIG. 16, the start time and the end time have relatively high priority, and therefore the size of the square corresponding to the start time and the end time. Since the priority of health care time is relatively low, the size of the mass corresponding to the health care time is reduced.

As a result, the size of the square of the action item having a higher priority becomes larger, so that the influence of the square having a higher priority becomes larger when the characteristics of the action image 403 are analyzed. That is, by assigning a high priority action item to a large area, it is possible to generate the action image 403 that enables analysis in consideration of the priority. Further, the feature table to be extracted can be adjusted by setting the priority for each action item using the priority table 401.

The embodiments described above are examples for explaining the present invention, and are not intended to limit the scope of the present invention only to those embodiments. Those skilled in the art can implement the present invention in various other modes without departing from the scope of the present invention.

100... Data processing system 101... Work management device 102... Work daily acquisition processing unit 201... Image output device 202... Behavior image generation processing unit 301... Image analysis device 302... Image input processing unit 303... Image learning processing unit 304... Image Classification processing unit 305... Result output processing unit

Claims (11)

A data processing system for processing data on human behavior, comprising:
A data management unit that stores action data that accumulates measurement values obtained by measuring actions of predetermined action items by a person,
For each person, using a coloring rule that defines a color corresponding to the measured value for each action item, and an arrangement rule that determines where in the image the region colored by the coloring rule is arranged for each action item, An image output unit that generates an action image in which a region colored in a color corresponding to the measurement value of each action item included in the action data of a predetermined period is arranged,
An image analysis unit that learns a behavior image of a person and correct answer information indicating the nature and/or state of the person, and estimates the nature and/or state of the person from the behavior image of the person based on the learning result,
A data processing system having. The image analysis unit learns a behavior image of a person and correct answer information indicating the nature and/or state of the person to estimate the nature and/or state of the person based on the behavior image of the person. Generating an estimation algorithm, and estimating the property and/or state of the subject from the action image of the subject based on the estimation algorithm,
The data processing system according to claim 1. The arrangement rule is information in which a daily action item is assigned to each of a plurality of vertically and horizontally separated cells,
The image output unit generates an action image in which each cell is colored with a color corresponding to a measurement value of an action item assigned to the cell, based on data of a person for a predetermined period.
The data processing system according to claim 1. The coloring rule, for action items related to working hours, the areas of the same action items are defined as the same hue, and the darker color tone is set as the working time becomes longer.
The data processing system according to claim 1. The arrangement rule assigns daily action items to the plurality of cells so that dates are arranged in either the vertical direction or the horizontal direction and action items are arranged in the other direction.
The data processing system according to claim 3. Action items have priority,
The image output unit sets the area of the action item to a size according to the priority,
The data processing system according to claim 1. It further has an information presentation unit that displays information about the nature and/or condition of the subject.
The data processing system according to claim 1. The information presenting unit further displays an action image of the target person,
The data processing system according to claim 7. The learning is deep learning,
The data processing system according to claim 1. A data processing method for a computer device to process data relating to human behavior, comprising:
Stores behavior data that accumulates measurement values that measure the behavior of predetermined behavior items by people,
For each person, using a coloring rule that defines a color corresponding to the measured value for each action item, and an arrangement rule that determines where in the image the region colored by the coloring rule is arranged for each action item, Generates an action image in which regions colored in colors corresponding to the measured values of each action item included in the action data of a predetermined period are arranged,
Learning a person's action image and correct answer information indicating the person's property and/or state, and estimating the person's property and/or state from the person's action image based on the learning result.
A data processing method by which a computer executes a thing. A data processing program for processing data on human behavior,
Stores behavior data that accumulates measurement values that measure the behavior of predetermined behavior items by people,
For each person, using a coloring rule that defines a color corresponding to the measured value for each action item, and an arrangement rule that determines where in the image the region colored by the coloring rule is arranged for each action item, Generates an action image in which regions colored in colors corresponding to the measured values of each action item included in the action data of a predetermined period are arranged,
Learning a person's action image and correct answer information indicating the person's property and/or state, and estimating the person's property and/or state from the person's action image based on the learning result.
A data processing program that causes a computer to execute a thing.